{"pk":24341,"title":"Double Dissociations Emerge in a Flat Attractor Network","subtitle":null,"abstract":"Double dissociations were long considered a gold standard\nfor establishing functional modularity. However, Plaut (1995)\ndemonstrated that double dissociations could result without\nunderlying modularity. He damaged attractor networks with\nseparate orthographic and semantic layers (as well as a hid-\nden layer with feedback connections from semantics) that were\ntrained to map orthography to semantics. Damaging con-\nnections coming from either the orthographic layer or recur-\nrent semantic connections (to and from cleanup units) could\nboth yield double dissociations, with some models exhibit-\ning greater relative deficits for abstract words, and others for\nconcrete words. We investigated whether double dissocia-\ntions would emerge in a simpler attractor network with 2 sets\nof units (orthographic and semantic) and 2 layers of connec-\ntions (orthographic-to-semantic and recurrent semantic con-\nnections). Random damage to orthographic-semantic con-\nnections yielded double dissocations (some damaged mod-\nels showed stronger relative deficits for abstract words, while\nothers showed stronger relative deficits for concrete words).\nSemantic-semantic damage led only to concrete deficits. The\npresence of double dissociations given different degrees of\ndamage in each model reconfirm Plaut's (1995) findings in\nsimpler, “flat” attractor network (O'Connor, Cree, & McRae,\n2009), with less potential for modularity. The tendency for\nconcrete impairments given damage to the semantic attractor\nlevel is at once surprising and revealing; it demonstrates a di-\nvision of labor (and partial modularity) that emerges in this\nnetwork. We will discuss theoretical implications, as well as\nnext steps in this research program.","language":"eng","license":{"name":"","short_name":"","text":null,"url":""},"keywords":[{"word":"Cognitive Neuroscience; cognitive neuropsychology; Computational Modeling; Computational neuroscience"}],"section":"Papers with Poster Presentation","is_remote":true,"remote_url":"https://escholarship.org/uc/item/94f2t1kd","frozenauthors":[{"first_name":"Ihintza","middle_name":"","last_name":"Malharin","name_suffix":"","institution":"BCBL: Basque Center on Cognition, Brain and Language","department":""},{"first_name":"Simona","middle_name":"","last_name":"Mancini","name_suffix":"","institution":"Basque Center on Cognition, Brain and Language","department":""},{"first_name":"James","middle_name":"","last_name":"Magnuson","name_suffix":"","institution":"University of Connecticut","department":""}],"date_submitted":null,"date_accepted":null,"date_published":"2024-01-01T18:00:00Z","render_galley":null,"galleys":[{"label":"PDF","type":"pdf","path":"https://journalpub.escholarship.org/cognitivesciencesociety/article/24341/galley/13938/download/"},{"label":"PDF","type":"pdf","path":"https://journalpub.escholarship.org/cognitivesciencesociety/article/24341/galley/21108/download/"}]}